QSAR analysis of the time- and dose-dependent anti-inflammatory in vivo activity of substituted imidazo[1,2-a]pyridines using artificial neural networks

The quantitative analysis and physicochemical description of time- and dose-dependent in vivo drug effects is problematic as observed effects depend on both pharmacodynamics and pharmacokinetics. These factors depend in a different manner on the physicochemical properties of the investigated drugs. Obviously in vivo effects of drug series are governed by highly nonlinear relationships. As the function relating in vivo anti-inflammatory effects of the imidazo[1,2-a]pyridines observed in rats (carrageenan-induced rat paw edema test) to dose, time, and physicochemical properties could not be modeled explicitly, an attempt was made to train an Artificial Neural Network (ANN) to find the unknown nonlinear relationship between input variables and observed % effect values. The analysis extracted a maximum of information from available in vivo data and resulted in an ANN model that enabled the calculation of dose- and property-dependent anti-inflammatory activities of a set of 15 imidazopyridines.